fallow_extract/
lib.rs

1//! Parsing and extraction engine for fallow codebase intelligence.
2//!
3//! This crate handles all file parsing: JS/TS via Oxc, Vue/Svelte SFC extraction,
4//! Astro frontmatter, MDX import/export extraction, CSS Module class name extraction,
5//! HTML asset reference extraction, and incremental caching of parse results.
6
7#![warn(missing_docs)]
8#![cfg_attr(not(test), deny(clippy::disallowed_methods))]
9#![cfg_attr(
10    test,
11    allow(
12        clippy::unwrap_used,
13        clippy::expect_used,
14        reason = "tests use unwrap and expect to keep fixture setup concise"
15    )
16)]
17
18mod asset_url;
19pub mod astro;
20pub mod cache;
21pub(crate) mod complexity;
22pub mod css;
23pub mod css_classes;
24pub mod css_in_js;
25pub mod css_metrics;
26pub mod flags;
27pub mod glimmer;
28pub(crate) mod graphql;
29pub(crate) mod html;
30pub(crate) mod iconify;
31pub mod inventory;
32pub mod mdx;
33mod module_info;
34mod parse;
35pub mod sfc;
36pub mod sfc_css;
37mod sfc_props;
38mod sfc_template;
39pub mod similar_code;
40mod source_map;
41pub mod suppress;
42/// Tailwind CSS arbitrary-value detection.
43pub mod tailwind;
44pub(crate) mod template_complexity;
45mod template_expression_scan;
46mod template_usage;
47/// Visitor utilities for AST extraction.
48pub mod visitor;
49
50use std::path::Path;
51
52use rayon::prelude::*;
53
54use cache::CacheStore;
55use fallow_types::discover::{DiscoveredFile, FileId};
56
57pub use fallow_types::extract::{
58    AngularComponentFieldArrayTypeFact, AngularTemplateMemberAccessFact, AngularThisSpreadFact,
59    ClassHeritageInfo, ClassThisMemberAccessFact, ClassThisWholeObjectUseFact,
60    ComputedEnumKeyUseFact, DefaultImportWholeObjectUseFact, DynamicCustomElementRenderFact,
61    DynamicImportInfo, DynamicImportPattern, ExportInfo, ExportName, FactoryCallMemberAccessFact,
62    FactoryFnMemberAccessFact, FactoryFnWholeObjectFact, FactoryReturnExport,
63    FactoryReturnObjectPropertyAccessFact, FactoryReturnObjectShapeExport,
64    FluentChainMemberAccessFact, FluentChainNewMemberAccessFact, ImportInfo, ImportedName,
65    InstanceExportBindingFact, LocalTypeDeclaration, MemberAccess, MemberInfo, MemberKind,
66    ModuleInfo, ModuleLoadMechanism, ParseResult, PlaywrightFixtureAliasFact,
67    PlaywrightFixtureDefinitionFact, PlaywrightFixtureTypeFact, PlaywrightFixtureUseFact,
68    PublicSignatureTypeReference, ReExportInfo, RequireCallInfo, RequiredTypeMemberFact,
69    SemanticFact, SourceReadFailure, StringEnumMemberValueFact, TypeAliasSurfaceTargetFact,
70    TypeMemberTypeEntry, TypedPropertyMemberAccessFact, VisibilityTag, VitestModuleMockAction,
71    VitestModuleMockOperationFact, compute_line_offsets,
72};
73
74pub use astro::{
75    extract_astro_frontmatter, extract_astro_style_regions, extract_astro_template_regions,
76};
77pub use css::{
78    ThemeScan, ThemeTokenDef, extract_apply_tokens, extract_apply_tokens_located,
79    extract_css_module_exports, extract_css_var_reads_located, scan_theme_blocks,
80};
81pub use css_classes::{
82    MarkupClassScan, MarkupClassToken, is_edit_distance_one, is_typo_edit, scan_markup_class_tokens,
83};
84pub use css_in_js::{
85    ConsumerQuery, CssInJsObjectSheets, CssInJsToken, CssInJsTokenDef, CssInJsTokenOrigin,
86    TokenConsumerHit, css_in_js_consumer_scan, css_in_js_object_sheets, css_in_js_theme_consumers,
87    css_in_js_theme_token_defs, css_in_js_token_consumers, css_in_js_token_defs,
88    css_in_js_virtual_stylesheet, panda_style_value_consumers, panda_token_call_consumers,
89};
90pub use css_metrics::{compute_css_analytics, parse_css_color_rgb};
91pub use glimmer::{is_glimmer_file, strip_glimmer_templates};
92pub use mdx::{extract_mdx_statements, extract_mdx_statements_mapped};
93pub use sfc::{
94    SourceRegion, extract_sfc_scripts, extract_sfc_styles, extract_sfc_template_regions,
95    is_sfc_file,
96};
97pub use sfc_css::{
98    scoped_unused_classes, sfc_preprocessor_virtual_stylesheet, sfc_virtual_stylesheet,
99};
100pub use similar_code::extract_similar_code_functions;
101pub use source_map::ExtractionResult;
102pub use tailwind::{TailwindArbitraryUse, scan_tailwind_arbitrary_values};
103
104#[expect(
105    clippy::expect_used,
106    reason = "static regex patterns are hard-coded analyzer invariants covered by extraction tests"
107)]
108fn static_regex(pattern: &str) -> regex::Regex {
109    regex::Regex::new(pattern).expect("static regex pattern should compile")
110}
111
112pub use parse::parse_source_to_module;
113
114/// Leading UTF-8 byte order mark codepoint.
115///
116/// Windows editors (Notepad, older VS settings, some IDE plugins) emit a UTF-8
117/// BOM at the start of source files. fallow's contract is "UTF-8 with or
118/// without BOM; line offsets are computed against the post-BOM view; the BOM,
119/// if present on input, is preserved on output by `fallow fix`."
120const BOM_CHAR: char = '\u{FEFF}';
121// Small, cache-hot inputs are faster on one thread than through Rayon setup.
122// Larger file sets still use parallel parsing where parse work dominates.
123const PARALLEL_PARSE_FILE_THRESHOLD: usize = 32;
124
125/// Strip the leading UTF-8 BOM if present.
126///
127/// Called at every file-read entry point in this crate so the rest of the
128/// pipeline (content hash, `compute_line_offsets`, oxc parser, downstream
129/// analyses) sees a consistent post-BOM view. Mirrors the
130/// `fallow_config` layer (`config_writer.rs::BOM`) so config-shaped sources
131/// and source-code-shaped sources are processed symmetrically. See issue #475.
132#[must_use]
133fn strip_bom(source: &str) -> &str {
134    source.strip_prefix(BOM_CHAR).unwrap_or(source)
135}
136
137/// Parse all files, extracting imports and exports.
138///
139/// Small file sets use a sequential fast path to avoid parallel scheduling
140/// overhead; larger file sets use parallel extraction.
141/// Uses the cache to skip reparsing files whose content hasn't changed.
142///
143/// When `need_complexity` is true, per-function cyclomatic/cognitive complexity
144/// metrics are computed during parsing (needed by the `health` command).
145/// Pass `false` for dead-code analysis where complexity data is unused.
146pub fn parse_all_files(
147    files: &[DiscoveredFile],
148    cache: Option<&CacheStore>,
149    need_complexity: bool,
150) -> ParseResult {
151    let results: Vec<ParseFileResult> = if files.len() <= PARALLEL_PARSE_FILE_THRESHOLD {
152        files
153            .iter()
154            .map(|file| parse_single_file_cached(file, cache, need_complexity))
155            .collect()
156    } else {
157        files
158            .par_iter()
159            .map(|file| parse_single_file_cached(file, cache, need_complexity))
160            .collect()
161    };
162
163    let mut modules = Vec::with_capacity(results.len());
164    let mut read_failures = Vec::new();
165    let mut hits = 0usize;
166    let mut misses = 0usize;
167    let mut parse_cpu_nanos = 0u64;
168
169    for result in results {
170        hits += result.cache_hits;
171        misses += result.cache_misses;
172        parse_cpu_nanos = parse_cpu_nanos.saturating_add(result.parse_cpu_nanos);
173        if let Some(module) = result.module {
174            modules.push(module);
175        }
176        if let Some(failure) = result.read_failure {
177            read_failures.push(failure);
178        }
179    }
180
181    if hits > 0 || misses > 0 {
182        tracing::info!(
183            cache_hits = hits,
184            cache_misses = misses,
185            "incremental cache stats"
186        );
187    }
188
189    ParseResult {
190        modules,
191        read_failures,
192        cache_hits: hits,
193        cache_misses: misses,
194        parse_cpu_ms: parse_cpu_nanos as f64 / 1_000_000.0,
195    }
196}
197
198struct ParseFileResult {
199    module: Option<ModuleInfo>,
200    read_failure: Option<SourceReadFailure>,
201    cache_hits: usize,
202    cache_misses: usize,
203    parse_cpu_nanos: u64,
204}
205
206impl ParseFileResult {
207    fn cache_hit(module: ModuleInfo) -> Self {
208        Self {
209            module: Some(module),
210            read_failure: None,
211            cache_hits: 1,
212            cache_misses: 0,
213            parse_cpu_nanos: 0,
214        }
215    }
216
217    fn cache_miss(module: ModuleInfo, parse_cpu_nanos: u64) -> Self {
218        Self {
219            module: Some(module),
220            read_failure: None,
221            cache_hits: 0,
222            cache_misses: 1,
223            parse_cpu_nanos,
224        }
225    }
226
227    fn read_failure(file: &DiscoveredFile, error: &std::io::Error) -> Self {
228        Self {
229            module: None,
230            read_failure: Some(SourceReadFailure {
231                file_id: file.id,
232                path: file.path.clone(),
233                error: error.to_string(),
234            }),
235            cache_hits: 0,
236            cache_misses: 0,
237            parse_cpu_nanos: 0,
238        }
239    }
240}
241
242/// Parse a single file, consulting the cache first.
243///
244/// Cache validation strategy (fast path -> slow path):
245/// 1. Open the file so unreadable sources cannot use stale cached analysis
246/// 2. Read mtime + size from the open handle
247/// 3. If mtime+size match the cached entry -> cache hit, return immediately
248/// 4. If mtime+size differ -> read file, compute content hash
249/// 5. If content hash matches cached entry -> cache hit (file was `touch`ed but unchanged)
250/// 6. Otherwise -> cache miss, full parse
251fn parse_single_file_cached(
252    file: &DiscoveredFile,
253    cache: Option<&CacheStore>,
254    need_complexity: bool,
255) -> ParseFileResult {
256    let cached_by_path = cache.and_then(|store| store.get_by_path_only(&file.path));
257
258    if let Some(cached) = cached_by_path
259        && cached.file_size == file.size_bytes
260    {
261        let source_file = match std::fs::File::open(&file.path) {
262            Ok(source_file) => source_file,
263            Err(error) => return ParseFileResult::read_failure(file, &error),
264        };
265        if let Ok(metadata) = source_file.metadata()
266            && metadata.len() == cached.file_size
267        {
268            let fingerprint =
269                fallow_types::source_fingerprint::SourceFingerprint::from_metadata(&metadata);
270            if cached.source_fingerprint() == fingerprint
271                && fingerprint.has_known_mtime()
272                && (!need_complexity || !cached.complexity.is_empty())
273            {
274                return ParseFileResult::cache_hit(cache::cached_to_module_opts(
275                    cached,
276                    file.id,
277                    need_complexity,
278                ));
279            }
280        }
281    }
282
283    let raw = match std::fs::read_to_string(&file.path) {
284        Ok(raw) => raw,
285        Err(error) => return ParseFileResult::read_failure(file, &error),
286    };
287    let source = strip_bom(&raw);
288    let content_hash = xxhash_rust::xxh3::xxh3_64(source.as_bytes());
289
290    if let Some(cached) = cached_by_path
291        && cached.content_hash == content_hash
292        && (!need_complexity || !cached.complexity.is_empty())
293    {
294        return ParseFileResult::cache_hit(cache::cached_to_module_opts(
295            cached,
296            file.id,
297            need_complexity,
298        ));
299    }
300
301    let parse_start = std::time::Instant::now();
302    let module = parse_source_to_module(file.id, &file.path, source, content_hash, need_complexity);
303    let parse_cpu_nanos = u64::try_from(parse_start.elapsed().as_nanos()).unwrap_or(u64::MAX);
304    ParseFileResult::cache_miss(module, parse_cpu_nanos)
305}
306
307/// Parse a single file and extract module information (without complexity).
308#[must_use]
309pub fn parse_single_file(file: &DiscoveredFile) -> Option<ModuleInfo> {
310    let raw = std::fs::read_to_string(&file.path).ok()?;
311    let source = strip_bom(&raw);
312    let content_hash = xxhash_rust::xxh3::xxh3_64(source.as_bytes());
313    Some(parse_source_to_module(
314        file.id,
315        &file.path,
316        source,
317        content_hash,
318        false,
319    ))
320}
321
322/// Parse from in-memory content (for LSP, includes complexity).
323#[must_use]
324pub fn parse_from_content(file_id: FileId, path: &Path, content: &str) -> ModuleInfo {
325    let content = strip_bom(content);
326    let content_hash = xxhash_rust::xxh3::xxh3_64(content.as_bytes());
327    parse_source_to_module(file_id, path, content, content_hash, true)
328}
329
330#[cfg(all(test, not(miri)))]
331mod tests;
fallow_extract/lib.rs

fallow_extract/
lib.rs
